Summing philosophy?

[Cojo]

Ok, let's assume this:

You have a computerbased recordingsystem running your favorite software conected via a soundcard of your own choice and then summed together in the analog domain using your one of a kind mixer. Suddenly you run out of outputs and end up submixing in the digital domain!

How should you best do this submixes, and still get the best audioquality?

Let's say we have 8 physical outputs and we have 16 tracks. Is it better to:

A - Mix all tracks 2 and 2
B - Mix tracks containing hi freq together and tracks containing lo freq together
C - Mix tracks containing lo freq together with hi freq tracks
D - Mix tracks containing "backgrouds and less importent" together and leave the "front end stuff" on single tracks
E - Leave 7 tracks as single tracks and the last one with the rest
F - A combination of any of the above

Does this make any sens at all? I think that there must be "a best way" for a digital system to sum waveforms that will benefit the sound and maby put less strain to the computer. Am I totally wrong?

/Cojo

[thethrillfactor]

It really depends on the arrangement and if its a summing device or an actual mixer.

If its a mixer and you have 16 tracks that you have to get down to eight i would probably have the kick,snare,lead vocal,bass and stereo backgroundvocals by themselves for special outboard treatment.

To me these are the important elements of all popular music styles and they have to be stellar.

The last 2 outs i would submix the sources in the box unless something was really suffering and that i would process externally and retrack.

[John Suitcase]

I have a similar situation, but what I do is sub the guitars together, the drums together, the vocals together, and my effects returns together. Then I'll use some compression/limiting on each subgroup. It sometimes help glue things together without making it too mooshy.

So the drums may pump a little when the kick and snare hit, but the rest of the mix doesn't. The guitars sort of fill each other in without stepping over the other elements.

My only thing is whether it'd be better to send the overheads out to the "effects" bus, so that I can slam the rest of the drumkit harder.

That and that I'd really like to have another buss to send the bass guitar and kick to.

If you're doing any group compression stuff, this is really the only way to go...

[subspace]

Mixdown to 2 outputs and use the other six for cowbell mults.

[John Suitcase]

Originally Posted by subspace

Mixdown to 2 outputs and use the other six for cowbell mults.

Ditto that.

Except I think you should probably just sum to mono, then use the other 7 channels for true surround.

Cowbells all around...

[zetterstroem]

[TornadoTed]

Originally Posted by thethrillfactor

It really depends on the arrangement and if its a summing device or an actual mixer.

If its a mixer and you have 16 tracks that you have to get down to eight i would probably have the kick,snare,lead vocal,bass and stereo backgroundvocals by themselves for special outboard treatment.

To me these are the important elements of all popular music styles and they have to be stellar.

The last 2 outs i would submix the sources in the box unless something was really suffering and that i would process externally and retrack.

I couldn't agree more, that is exactly to the letter how I go about things. I have 16 inputs but the kick, snare, bass and lead vocals all have their own seperate channel.

[Cojo]

Originally Posted by thethrillfactor

CUT-->
If its a mixer and you have 16 tracks that you have to get down to eight i would probably have the kick,snare,lead vocal,bass and stereo backgroundvocals by themselves for special outboard treatment.

To me these are the important elements of all popular music styles and they have to be stellar.

The last 2 outs i would submix the sources in the box unless something was really suffering and that i would process externally and retrack.<--CUT

Yeah, this is what I do today. The convienient way of mixing (drums on the drumbus, rythmguitar here and lead there etc) may be good in the analog domain but when it comes to digital the bits and pieces don't mix so well all the times.

But don't you worry cause I've been thinkin' that maby this isn't the best way (sonicly) to do it?...

I've posted some audio and graphix to illustrate what I meen. What I've done is that I've recorded some sinewaves (C0, C#0, C8, C#8) in different combinations and the result is, hmmm... how should I say this... most terrifying!

Ok, some explanations to the files (all recorded in 44.1kHz & 24bits):

1. Pure sinewave C0
2. Pure sinewave C8 (I was strucked by the way a sinewave in this freq looked i digital media)
3. Mix of C0 and C#0 (Now it's getting scary. The two waves start modulating each other and note the increase in audiolevel! It actually clips.)
4. Mix of C8 and C#8 (Pretty much the same phenonomen as in nr.3 but without clipping.)
5. Mix of C0 and C8 (Here the hi freq actually rides the wave. But no clipping and the two waves are actually pretty intact.)

This is whats been bothering my brain lately!?
Mayby the obvious isn't so obvious any more? Sleep tight!

/Cojo

[Cojo]

Here's the rest of the files!

[Paul Frindle]

Originally Posted by Cojo

Ok, some explanations to the files (all recorded in 44.1kHz & 24bits):

1. Pure sinewave C0
2. Pure sinewave C8 (I was strucked by the way a sinewave in this freq looked i digital media)
3. Mix of C0 and C#0 (Now it's getting scary. The two waves start modulating each other and note the increase in audiolevel! It actually clips.)
4. Mix of C8 and C#8 (Pretty much the same phenonomen as in nr.3 but without clipping.)
5. Mix of C0 and C8 (Here the hi freq actually rides the wave. But no clipping and the two waves are actually pretty intact.)

This is whats been bothering my brain lately!?
Mayby the obvious isn't so obvious any more? Sleep tight!

/Cojo

If you mix together 2 waveforms at any level above -6dB (50%) they will cause clipping because at some point the peaks of the waves will coincide to add to something greater than + or -1.

For instance if you C0 and Csharp0 are both below -6dB the clipping shouldn't happen. If it does you have some extra limiter in the signal path down line or the system is broken.

Also be aware that you waveform when viewed within the DAW is showing you undecoded sample values - NOT the signal that will finally come out of your DAC. This is terrribly important as it doesn't give you anything like a true representation of the decoded signal and it therefore encourages people to believe all sorts of confusing things :-)

[FMNYC]

i have only eight analog outs available to me in my system.
i used to do submixes until i had everything down to just music and vocals separately but what works for me is to "submix" by routing drums to a stereo pair, guitars to another stereo pair, keys and strings to another... you get the idea.

this has given me the bes results sonically.
i use an API 8200a as my summing device and it seems to works best this way, instead of submixing.

my humble rookie contribution...

peace!

FM

FM has traveled in time just to meet Sarah Connor.

[raal]

Originally Posted by Paul Frindle

If you mix together 2 waveforms at any level above -6dB (50%) they will cause clipping because at some point the peaks of the waves will coincide to add to something greater than + or -1.

For instance if you C0 and Csharp0 are both below -6dB the clipping shouldn't happen. If it does you have some extra limiter in the signal path down line or the system is broken.

Also be aware that you waveform when viewed within the DAW is showing you undecoded sample values - NOT the signal that will finally come out of your DAC. This is terrribly important as it doesn't give you anything like a true representation of the decoded signal and it therefore encourages people to believe all sorts of confusing things :-)

mr. frindle and others shed much light on this topic here: http://recforums.prosoundweb.com/ind...sg/4918/0/96/0

if i remember correctly, it gets pertinent starting around page 7. very enlightening.

i would like to take this opportunity to thank mr. frindle for having taken the time to explain on that thread.

[Cojo]

Originally Posted by Paul Frindle

CUT-->
Also be aware that you waveform when viewed within the DAW is showing you undecoded sample values - NOT the signal that will finally come out of your DAC. This is terrribly important as it doesn't give you anything like a true representation of the decoded signal and it therefore encourages people to believe all sorts of confusing things :-)<--CUT

Aha, this was new to me... and can also explain why better convertes sounds better! You learn something new everyday.

/Cojo

[Cojo]

Originally Posted by raal

mr. frindle and others shed much light on this topic here: http://recforums.prosoundweb.com/ind...sg/4918/0/96/0

if i remember correctly, it gets pertinent starting around page 7. very enlightening.

i would like to take this opportunity to thank mr. frindle for having taken the time to explain on that thread.

Thanks for posting this link! And thanks Paul for the rest of the info! Verry good reading indead. I put a big star in your book. thumbsup

However, I'm still wondering if the computer will do a better job calculating those summing busses if I only sum two sources instead of 8? Or if it's better summed if Lo and Hi content is summed rather then Lo and Lo?

I have a feeling... you more you know, you more question you have.

/Cojo

[Paul Frindle]

Originally Posted by Cojo

Thanks for posting this link! And thanks Paul for the rest of the info! Verry good reading indead. I put a big star in your book. thumbsup

However, I'm still wondering if the computer will do a better job calculating those summing busses if I only sum two sources instead of 8? Or if it's better summed if Lo and Hi content is summed rather then Lo and Lo?

I have a feeling... you more you know, you more question you have.

/Cojo

Thanks for the thanks :-) The long thread on PSW is worth a read since lots of this stuff is discussed there - but it does take some time to get through.

One thing that computers do with almost perfect accuracy is add stuff together, but the problem comes with the fact that computers add undecoded sample values - not signal. So it's possible the create samples which cause illegal signals during decoding even though they were apparently ok at the time of the mix. The other problem is that the DAWs have no headroom and the people who use them modulate almost flat out.

Basically it should not matter what you sum to what provided that you don't cause sample value overs (metering red lights) or signal overs (reconstructed errors out of the DAC). The first is easy to see cos all DAWs show sample value, but the reconstruction levels are not shown on the DAW. The limiter we make includes special processing to show you the reconstruction levels (and even fix them dynamically) so as to avoid surprises on DACs downline - either your's or the customers. But failing that the best, cheapest and surest thing to do is simply reduce all the levels by 3 to 6dB - like that there is no chance of causing overs of any sort. You can then let the mastering guy increase it to the required levels for fashion - as (hopefully) he has the kit to so this accurately and safely. There's an explanation of this which I had included in the limiter blurb linked below

http://www.sonyoxford.co.uk/pub/plug...ech_Detail.htm

BTW your clipping sinewaves demo (C0 and C#0) is a good illustration of the really crucial headroom issue. For instance in your tones that cause overs, if you had sent them each out of a DAC both would have been ok because they don't get too big - when you sum them in the analogue domain everything is still ok cos the analogue mixer has headroom (i.e. the metered maximum level is less than it can stand before saturation - so it doesn't crap out when adding the signals together).

To achieve the same in your DAW you need only to reduce the levels in the digital domain so that they don't saturate when added up :-) The problem with the DAW is that it has a metering calibrated to read max level at the very point of saturation (no headroom) - this is a mistake in all common DAW systems which have become the norm.

You can avoid the whole darned problem by simply reducing levels at the begining of every channel (each track into the mixer) and aiming for -6 to -10dB on the meters rather than flat out - and putting back the level again at the final output of the mixer. You can afford to do this as a 24bit system has around 140dB of dynamic range - loads more than you will ever need. Give it a try you might be really surprised - and you might avoid the analogue summing which is less accurate and has problems of it's own :-)

[raal]

Originally Posted by Paul Frindle

You can avoid the whole darned problem by simply reducing levels at the begining of every channel (each track into the mixer) and aiming for -6 to -10dB on the meters rather than flat out - and putting back the level again at the final output of the mixer.< snip> Give it a try you might be really surprised - and you might avoid the analogue summing which is less accurate and has problems of it's own :-)

i gave it a try and i was really surprised... paul frindle should have a statue in the audio walk of fame.

for those interested in the subject, go here and download the white paper http://www.tllabs.com/index.php?opti...d=20&Itemid=62 i bought the meter. i'm happy. still like analog stuff though.

[Saudade]

Mr Frindle!

I have also tried your method in SX and heard a huge difference. I used to believe software summing somehow causes the "murkiness/cloudiness". Back then, it seems whatever EQ/Compr moves I make I just couldn't hear them working the way is supposed to. Even when resorting to huge boosts and cuts does not yield anything close to the desired effect, only screws up the sound more. But just backing off all the pre-plugin track levels by 6 or more I can hear the cloudiness is gone. Now all my EQ moves are very apparent and even each minor adjustments I can hear the difference. My confidence balloons!

Gosh this is starting to sound like some cheesy slimming advertisment customer testimonial

Just a question Mr Frindle, does the "4x oversampling" feature offered in some software limiters strapped on the "2-buss" help to detect intersample (signal)peaks? Is ensuring the sample value does not exceed -3 the best way in the inavailability of mastering services (budget projects)?

Hope to see you around GS more often and perhaps come on as guest moderator sometime?

There are a lot of paranoid people like me who are (excessively) obsessed about what 1s and 0s are doing to their sound....

[John Suitcase]

Interesting. Are you saying that during tracking, I should aim for -6db rather than trying to approach 0db? Or are you saying that once tracked, I should pull those levels back so they have some headroom, rather than pushing the fader up farther? Or both?

I've also wondered about the costs of analogue summing, as I prefer the results, but I know logically that I'm adding noise, plus I'm adding a whole D/A A/D step that must degrade the sound somewhat.

If summing ITB can be made to work without dither/saturation/clipping errors, then it should be superior to the whole analogue summing process...

[raal]

Originally Posted by John Suitcase

Are you saying that during tracking, I should aim for -6db rather than trying to approach 0db? Or are you saying that once tracked, I should pull those levels back so they have some headroom, rather than pushing the fader up farther? Or both?

someone correct me if i'm wrong, but you can record relatively hot (no overs) and then insert a trim plug. i actually go -6 to -8db to be extra safe. it's best to insert an intersample peak meter between plugs also to check, but as mr. frindle said, if you go -6 you're probably out of the woods in most cases. as i understand it the problem is not usually in the DAW's summing - it's when cheapo converters to try to reconstruct waves with intersample peaks that they are not able to handle, and with some plug ins. the white paper i mentioned earlier goes into this.

[Cojo]

Originally Posted by Paul Frindle

CUT-->Thanks for the thanks :-)<--CUT

You're wellcome!

Originally Posted by Paul Frindle

CUT-->
BTW your clipping sinewaves demo (C0 and C#0) is a good illustration of the really crucial headroom issue. For instance in your tones that cause overs, if you had sent them each out of a DAC both would have been ok because they don't get too big - when you sum them in the analogue domain everything is still ok cos the analogue mixer has headroom (i.e. the metered maximum level is less than it can stand before saturation - so it doesn't crap out when adding the signals together).

To achieve the same in your DAW you need only to reduce the levels in the digital domain so that they don't saturate when added up :-) The problem with the DAW is that it has a metering calibrated to read max level at the very point of saturation (no headroom) - this is a mistake in all common DAW systems which have become the norm.

You can avoid the whole darned problem by simply reducing levels at the begining of every channel (each track into the mixer) and aiming for -6 to -10dB on the meters rather than flat out - and putting back the level again at the final output of the mixer. You can afford to do this as a 24bit system has around 140dB of dynamic range - loads more than you will ever need. Give it a try you might be really surprised - and you might avoid the analogue summing which is less accurate and has problems of it's own :-)<--CUT

Finally I'm totally satisfied! Now I know what I wanted to know... and a bit more!

Thanks all for helping!

/Cojo

[Paul Frindle]

Originally Posted by John Suitcase

Interesting. Are you saying that during tracking, I should aim for -6db rather than trying to approach 0db? Or are you saying that once tracked, I should pull those levels back so they have some headroom, rather than pushing the fader up farther? Or both?

I've also wondered about the costs of analogue summing, as I prefer the results, but I know logically that I'm adding noise, plus I'm adding a whole D/A A/D step that must degrade the sound somewhat.

If summing ITB can be made to work without dither/saturation/clipping errors, then it should be superior to the whole analogue summing process...

You can track up to peak (of course avoiding any sample value overs) and reduce the level of the track back into the mix application using some digital trim plug or some such. This is a bit better than reducing recording levels at the ADC input as noise may intrude depending on how good the converter actually is. But any way you get the level down is better than leaving it near flat out when mixing :-)

[Paul Frindle]

Originally Posted by Saudade

Mr Frindle!


Just a question Mr Frindle, does the "4x oversampling" feature offered in some software limiters strapped on the "2-buss" help to detect intersample (signal)peaks? Is ensuring the sample value does not exceed -3 the best way in the inavailability of mastering services (budget projects)?

Hope to see you around GS more often and perhaps come on as guest moderator sometime?

There are a lot of paranoid people like me who are (excessively) obsessed about what 1s and 0s are doing to their sound....

The answer is that they may do - but it's not at all certain they actually do.

It is important to understand that running at higher sample rates does NOT avoid intersample peaks caused by processing - unless you sacrifice the extra freq range in the converters themselves. I.e. running the whole thing at say 192KHz but restricting the bandwidth of the signal to 20KHz still by filtering it in the actual DAC. As ever - the problem with simply running at higher sample rates is what exactly is IN the freq ranges you can't hear and don't need - it's not necessarily useful stuff, despite what the marketeers would have you believe :-(

So in order to avoid the intersample peaks in a compressor you need to reconstruct the actual output waveform levels the DAC will yield in the output. Just upsampling and running at 4FS does not get you that automatically - unless a full reconstruction at the output sample rate is done within the compressor sidechain as well.

Some reasons why this may not be done is that the filtering required is expensive in processing, causes a potential latency problem for fast compressor/limiters and makes the compressor less accurate in terms of absolute level wrt freq. And of course the sample value metering in the DAW may give confusing results for the user if the slightest difference occurs between settings and the meter readings (i.e. people aiming at the red light etc). And one should not forget that the action and character of the compression with different signals would be modified - for instance under some conditions it would seem to overcompress in ways that might annoy.

When I designed the Oxford limiter I actually treated the reconstruction metering (and the extra processing to dynamically correct the reconstruction errors) as a totally separate process. However it's quite costly - people using host LE or RTAS versions will notice the the processing load increases dramatically when the meter is switched to recon or the auto comp function is used - and that's with none of the processing actually running signal at higher sample rates. You do not have to upsample the signal itself to represent the reconstructed level - it's all done by other means.

As for spending more time on GS - in the past it's really been a question of available spare time. Thanks for the encouragement for the future though :-)

[Paul Frindle]

Originally Posted by raal

someone correct me if i'm wrong, but you can record relatively hot (no overs) and then insert a trim plug. i actually go -6 to -8db to be extra safe. it's best to insert an intersample peak meter between plugs also to check, but as mr. frindle said, if you go -6 you're probably out of the woods in most cases. as i understand it the problem is not usually in the DAW's summing - it's when cheapo converters to try to reconstruct waves with intersample peaks that they are not able to handle, and with some plug ins. the white paper i mentioned earlier goes into this.

Sorry Raal - I just repeated your exact answer before reading yours.

Yes -6dB will avoid all intersample peaks (as far as I can tell) however bad they may be. But please be aware that other processes can also increase levels by as much as 6dB as well - without actually making it sound ANY louder at all - it may even get quieter! The only saving grace is that these sorts of issues will bring on the red light (unlike intersample peaks) - but who wants to mix whilst worrying about overs the whole time? My advice is reduce the level into your mix app by as much as 20dB - you can then create your sounds with real freedom - just like the old analogue apps that operated 20dB below saturation :-)

BTW for those that need to aim for absolute maximum loudness to satisfy the popular market, there are messages in all this as to how you can actually manipulate the programme itself to be louder than other mixes - that though broadly similar in sound and feel, put red lights on much much earlier. If you read the PSW thread at the point where I talk about the effect of filtering a square wave the clues are there - clipping and overdriving stuff may make things louder individually, but it does not necessarily get you a final mix that sounds louder in the end at all! Fascinating stuff :-)

[juicemaster1500]

Originally Posted by Paul Frindle

One thing that computers do with almost perfect accuracy is add stuff together, but the problem comes with the fact that computers add undecoded sample values - not signal. So it's possible the create samples which cause illegal signals during decoding even though they were apparently ok at the time of the mix. The other problem is that the DAWs have no headroom and the people who use them modulate almost flat out.

Basically it should not matter what you sum to what provided that you don't cause sample value overs (metering red lights) or signal overs (reconstructed errors out of the DAC).

Does this apply to floating point DAWs as well?

Surely there's no illegal signals within the floating point DAW (no distortion going over - 0 dbs) yet summing to a master bus or other buses doesn't sound very good to my ears (I'm not overloading the DAC btw, I keep a good margin as to not get intersample peaks).

Btw, I use the Sony plug-ins for TC Powercore (and they rock), I've noticed that the Sony plug-in overload meters go off when other plug-in meters doesn't, does this mean that the Sony plug-in metering detects intersample peaks?

[Paul Frindle]

Originally Posted by juicemaster1500

Does this apply to floating point DAWs as well?

Surely there's no illegal signals within the floating point DAW (no distortion going over - 0 dbs) yet summing to a master bus or other buses doesn't sound very good to my ears (I'm not overloading the DAC btw, I keep a good margin as to not get intersample peaks).

The intersample peak situation is independent of any scaling or math respresentation since it is not dependent on sample overload conditions.

But talking of head room, whilst a float application 'may' allow you to have signals beyond the red light level and 'might' allow you to recover them (almost completely) by reducing down line, this situation is no different from a fixed mixer when you avoid the red lights. If the red lights are out - no sample value clip has happened on either system - so no difference should occur between them.
But the real thing to consider is when anything needs to have reference to the real world - the float system is completely free of scaling - but the music isn't! It must be represented in the real world at the end of the day in the DAC - but also the internal apps need reference to the real world too in order to produce the right effect. So if you think about things like compressors/limiters, harmonic generation in fact any process that has level dependent behaviour - they all must construct a reference of scaling with reality. For instance, in some of the plugs we do where math saturation is actually part of the wanted behaviour we actually deliberately limit the internal math to +/-1 to get the intended result - and of course make absolutely sure that apps are precisely the same between float and fixed point types :-)
Despite what people believe (and how it at first looks), the float mixer does not avoid scaling and level representation issues - all it does is provide a method to recover from sample overloads that happen ONLY within the mixer interfaces themselves.

Btw, I use the Sony plug-ins for TC Powercore (and they rock), I've noticed that the Sony plug-in overload meters go off when other plug-in meters doesn't, does this mean that the Sony plug-in metering detects intersample peaks?

No - only the Oxford limiter output meter will display intersample peaks. All the rest are simple sample value - it would cost too much processing to do otherwise.

Now this is a lovely example of scaling wrt the real world signal. The TC system uses fixed point processing - so although your host app is float the plugs on the PC expansion processing must receive and pass on fixed point signals. So in this case a representation of the float values must be made in the fixed point domain (and vice versus). In fixed point two's comp math the value -1 can be represented exactly, but +1 cannot! This is because a real code (sample value) must be designated as zero - (i.e. nothing, off etc) - and we only have an even number of codes, therefore one direction must end up with one more total count than the other. In fixed point this means we can only reach one count less than +1 at whatever the data width is in use.
Since the red lights on most DAWs come on at exactly flat out (no more, no less), it is possible that this translation leads to a one count difference between what the fixed point thinks is flat out and the float meters think is flat out.
For 24 bit this isn't very noticeable cos one count is not very big - but if for instance the data were represented as 16bit fixed the amount we would lose from the real value of +1 would be much bigger in real terms (256 times bigger in fact, which would mean around 0.00013dB error for a flat out signal). And it would also mean that a 16bit signal would clip 0.00013dB before the float would think it was actually at maximum - or the level (gain) from the fixed point would be 0.00013dB less than the floating mixer if we aligned the gains clip points exactly. Who cares about such a small error you might ask?
Well for instance there was one version of ProTools that had 16bit metering although the rest of the app was 24bits - it posed quite a problem to make some plugs put the red light on when the user would expect it to be on - simply because the vanishingly small difference in level that puts the light on or off is so crazily important to the user, who is using minute fractions of a dB to make judgments on levels - because of this popular obsession with maximum loudness and red lights.
For instance, who in this day and age would trust something like a limiter that could never put the red light on - even though it reached -0.00013dB of flat out!! :-(
The clip light has become everyone's reference level and this is a massive limitation to people - though they rarely realise it!!

[raal]

Originally Posted by Paul Frindle

My advice is reduce the level into your mix app by as much as 20dB - you can then create your sounds with real freedom - just like the old analogue apps that operated 20dB below saturation :-)

wow! 20 db? as in inserting a trim plug and bringing the individual channel down 20db?! i usually bring things down to -8db (after tracking). and thank you for all your replys on this thread sir. much appreciated by many of us.

[juicemaster1500]

Thanks for the detailed post, Paul Frindle.

The next question is;

when will we see the rest of the Sony Oxford line-up on TC Powercore?

[juicemaster1500]

Originally Posted by raal

wow! 20 db? as in inserting a trim plug and bringing the individual channel down 20db?!

I have crazy respect for mr. Frindle and the rockin' Sony Oxford plug-in line, but I wouldn't agree with this advice.

Hitting a plug-in with to low a signal doesn't sound right, and I can prove it

I've tested a lot of eq plug-ins for this, and they all have sort of a 'sweet spot'.

Even with the Sony Eq, I hit one with a signal -24db and another with a signal -12 db, made identical settings, same programme material, gained the lower amplitude track to compensate (with a waves q1 in bypass) phase reversed it, and in one instance the difference between the two channels was as much as -55 db! With flat settings the difference would be something like about a -105db's down, probably because there's a dither signal in the Sony algorithm, which would be 12 db's louder on the louder track.

This proves that there are differences in sound depending on how loud a signal a plug-in receives, even well below the threshold of intersample peaks.

I guess it's sort of a balancing act, avoiding intersample peaks and keeping a reasonable level for plug-ins to work with.

[Paul Frindle]

Originally Posted by juicemaster1500

Thanks for the detailed post, Paul Frindle.

The next question is;

when will we see the rest of the Sony Oxford line-up on TC Powercore?

This is a vexing issue as we care greatly for the PowerCore users and regret the disappointment it has caused in that fraternity. There is no intention of not providing powercore plugs. The honest answer is (sadly) a lack of resources.

The design team here at Oxford consists of only two people; myself who conceives the products, does the research, makes the processing and writes the operation manuals and blurb - and a programming expert who codes all these applications for the platforms, designs the GUI and user interfaces and makes all the great number of various instantiations.

We used to have one extra guy who conentrated on porting the plugs to Powercore, but he left more than a year ago and the various attempts to replace him have been unsuccessful so far.

Servicing the PT market (and distribution chain) with it's yearly host S/W updates, new H/W and system additions and preparing support stuff for 4 or 5 shows each year for the sales guys, uses just about all of our technical resources. For example, when we revamped the plugs to use iLok we had to re-build every instantiation of every plug application. Just testing them all again took me 5 months of continuous effort. This held up the design of the reverb which in the end took a total of 2 years work! The PT market provides the the much larger portion of our income and being a very small department of around 15 people we cannot financially afford to let it fall behind.

[Paul Frindle]

Originally Posted by juicemaster1500

I have crazy respect for mr. Frindle and the rockin' Sony Oxford plug-in line, but I wouldn't agree with this advice.

Hitting a plug-in with to low a signal doesn't sound right, and I can prove it

I've tested a lot of eq plug-ins for this, and they all have sort of a 'sweet spot'.

Even with the Sony Eq, I hit one with a signal -24db and another with a signal -12 db, made identical settings, same programme material, gained the lower amplitude track to compensate (with a waves q1 in bypass) phase reversed it, and in one instance the difference between the two channels was as much as -55 db!

The difference is too big and not the result of Oxford EQ plug inaccuracy. There is no 'sweet spot' with the Oxford EQ.

What is the waves 'q1'?

Not all plugs are of equal quality and there may indeed be examples of plugs around that may perform differently (or can't reach the required settings) at lower levels (particularly dynamics stuff). Most people's plugs will be designed to operate at max level cos that is what the industry is doing en masse. Obviously it's all a question of balancing advantage and disadvantage.

My advice is only technically based refering the the plugs we have made and the experience I have gained. In the end you must do what sounds right to you and what you are most comfortable with. It isn't my place to dictate or criticise in any way and I certainly wouldn't expect that anything I said here would change the world.

If there are even only a few people who benefit from the advice, it makes the effort worthwhile :-)

BTW - you don;t need to lose track input level to avoid intersample peaks, -6dB at the output is sufficient. The -20dB advice is only for providing the degree of freedom to experiment without clipping that an analogue system would normally have.